How to improve gemini-1.5-flash output accuracy on images

Gemini API
,
1

I am using gemini 1.5 flash for my project in which it will be given an image in jpg format and need to summarize it. The image contains handdrawn images with lenght breadth and other information. you can consider the image as a blueprint (handdrawn).

2

Hi @Nikhil_540

You can give a few shot prompts to the model or you can fine-tune the model for your handdrawn images.

Thanks

3

few short prompting works well and wll work with text problems but it would be daunting to few shot blueprints and i need to improve the accuracy using only 4 images so no finetuning

4

Try to break down into steps what you want the model to pay attention to within the image, what you want the model to do (process) with that information and how you want the model to output it.